Skip to content

Latest commit

 

History

History
40 lines (29 loc) · 1.8 KB

README.md

File metadata and controls

40 lines (29 loc) · 1.8 KB

github url github license badge Research Software Directory CI status

Echoframe tensor core beamformer

Requirements

Software Minimum version
CUDA 11.0
CMake 3.20
Hardware Type
GPU NVIDIA GPU with 1-bit tensor cores (Turing-generation or later)

Installation

This project uses CMake. To build the library:

git clone https://github.com/nlesc-recruit/echoframe-beamformer
cd echoframe-beamformer
cmake -S . -B build
make -C build

This will create executables and a libtcbf library in the build directory.

Usage

Prepare A matrix

The A matrix must be preprocessed before beamforming. Example commandline:

echoframe-tcbf-prepare-a-matrix --a_matrix_in A_matrix.bin --a_matrix_out A_matrix_preprocessed.bin --pixels 4096 --samples 16384

Standalone beamformer

The standalone beamformer executable is echoframe-beamformer. It can read a preprocessed A matrix and raw RF from disk, process it on the GPU, and store the BF back to disk. Example commandline:

echoframe-tcbf-standalone --a_matrix A_matrix_preprocessed.bin --rf RF.bin --bf BF.bin --pixels 4096 --frames 8192 --samples 16384